Abstract 85: Solutions for Selective Loss to Follow-Up: Application to HIV-Associated Cancer Cohorts in Malawi

Abstract

Purpose: Survival estimates among HIV+ cancer patients in sub-Saharan Africa (SSA) are limited and often biased due to high rates of loss to follow-up (LTFU). We explored the application of inverse probability of censoring weights (IPCW) to routinely collected HIV cohort data to correct for survival estimation bias introduced by LTFU, an approach which has been successfully utilized in high-income countries. Methods: We estimated overall survival (OS) among HIV+ incident cancer cases enrolled between 2000 and 2010 in Malawi. Patients were identified by utilizing probabilistic linkage of two large HIV cohorts to the national cancer registry and has previously been published. LTFU was defined as missed clinic appointment with unsuccessful tracing for ≥180 days. OS was estimated using unweighted and weighted methods. Sensitivity analysis using best- and worst-case scenarios assessed the robustness of estimates from the IPCW approach. Results: Among 883 cases, median age was 35 years (IQR: 30-41), and the most common cancers were Kaposi sarcoma (89%) and cervical cancer (5%). Median follow-up time was 5 years (IQR: 1-6), and 257 (29%) were LTFU. Males (RR: 1.69, CI: 1.37-2.07) and HIV stage IV patients (RR: 2.24, CI: 1.74-2.89) showed increased risk of LTFU. The 2-year OS were 91.1% and 89.6% for weighted and unweighted methods respectively. Sensitivity analysis estimated best and worst case 2-year OS at 91.5% and 77.4%, respectively. Conclusion: IPCW method did not appear to significantly correct bias introduced by LTFU, likely yielding continued overestimation of OS. This may be due to limited covariate availability in routinely collected cohort data in SSA to generate appropriate censoring weights or perhaps because of incorrect cancer diagnoses without pathologic confirmation in our setting. In SSA, where comprehensive patient tracing among LTFU groups can be impractical, intensive tracing of representative samples of LTFU patients is likely needed to accurately estimate survival. Citation Format: Evaristar Kudowa, Jess Edwards, Marie Josephe Horner, Steady Chasimpha, Dzamalala, Maganizo Chagomerana. Solutions for Selective Loss to Follow-Up: Application to HIV-Associated Cancer Cohorts in Malawi [abstract]. In: Proceedings of the 9th Annual Symposium on Global Cancer Research; Global Cancer Research and Control: Looking Back and Charting a Path Forward; 2021 Mar 10-11. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2021;30(7 Suppl):Abstract nr 85.